Abstract
This paper studies a risk-averse newsvendor pricing model with limited demand information under the conditional value-at-risk (CVaR) criterion. The paper uses a max–min approach and the objective is to maximize the lower bound on the CVaR of the loss, i.e., the negative of profit in the worst possible distribution case. The paper analyzes the optimal ordering and pricing decisions under both multiplicative and additive demand models, identifies the optimality conditions of the lower bound on the CVaR of loss, and obtains the implicit solutions for the optimal price and order quantity. Furthermore, the paper analyzes the sensitivity of optimal solutions with respect to the degree of risk aversion.
Similar content being viewed by others
References
Adhikary K, Roy J, Kar S (2018) A distribution-free newsboy problem with fuzzy-random demand. Int J Manag Sci Eng Manag 13(3):200–208
Agrawal V, Seshadri S (2000) Impact of uncertainty and risk aversion on price and order quantity in the newsvendor problem. Manuf Serv Oper Manag 2(4):410–422
Artzner P, Delbaen F, Eber JM, Heath D (1997) Thinking coherently. Risk 10(11):68–71
Artzner P, Delbaen F, Eber JM, Heath D (1999) Coherent measures of risk. Math Finance 9(3):203–228
Bertsimas D, Thiele A (2006) Robust and data-driven optimization: modern decision making under uncertainty. In: Models, Methods, and applications for innovative decision making, pp. 95–122
Ceryan O, Sahin O, Duenyas I (2013) Dynamic pricing of substitutable products in the presence of capacity flexibility. Manuf Serv Oper Manag 15(1):86–101
Chan LMA, Shen ZJM, Simchi-Levi D, Swann J (2004) Coordination of pricing and inventory decisions: a survey and classification. In: Handbook of quantitative supply chain analysis: modeling in the E-business Era, pp 335–392. Kluwer Academic Publishers, Boston
Chen YF, Xu M, Zhang ZG (2009) A risk-averse newsvendor model under the CVaR criterion. Oper Res 57(4):1040–1044
Dai J, Meng W (2015) A risk-averse newsvendor model under marketing-dependency and price-dependency. Int J Prod Econ 160:220–229
Dell M, Fredman C (1999) Direct from dell: strategies that revolutionized an industry. Harper Business, New York
DeYong GD (2020) The price-setting newsvendor: review and extensions. Int J Prod Res 58(6):1776–1804
Elmaghraby W, Keskinocak P (2003) Dynamic pricing in the presence of inventory considerations: research overview, current practices, and future directions. Manag Sci 49:1287–1309
Feng Q, Luo S, Shanthikumar JG (2020) Integrating dynamic pricing with inventory decisions under lost sales. Manag Sci 66(5):2232–2247
Fisher ML (1997) What is the right supply chain for your product? Harvard Business Review March–April, pp 83–93
Fu Q, Sim CK, Teo CP (2018) Profit sharing agreements in decentralized supply chains: a distributionally robust approach. Oper Res 66(2):500–513
Gallego G, Moon I (1993) The distribution free newsboy problem: review and extensions. J Oper Res Soc 44(8):825–834
Gotoh J, Takano Y (2007) Newsvendor solutions via conditional value-at-risk minimization. Eur J Oper Res 179(1):80–96
Gupta V, Cakanyildirim M (2016) A WTP-choice model: empirical validation, competitive and centralized pricing. Prod Oper Manag 25(11):1866–1884
Han Q, Du D, Zuluaga LF (2014) Technical note-a risk- and ambiguity-averse extension of the max–min newsvendor order formula. Oper Res 62(3):535–542
Kamburowski J (2014) The distribution-free newsboy problem under the worst-case and best-case scenarios. Eur J Oper Res 237(1):106–112
Liao Y, Banerjee A, Yan C (2011) A distribution-free newsvendor model with balking and lost sales penalty. Int J Prod Econ 133(1):224–227
Lu Y, Song M, Yang Y (2016) Joint inventory and pricing coordination with incomplete demand information. Prod Oper Manag 25(4):701–718
Minghui XU (2010) A price-setting newsvendor model under CVaR decision criterion with emergency procurement. J Syst Sci Syst Eng 19(1):85–104
Monahan GE, Petruzzi NC, Zhao W (2004) The dynamic pricing problem from a newsvendor’s perspective. Manuf Serv Oper Manag 6(1):73–91
Ogryczak WL, Ruszczynski A (2002) Dual stochastic dominance and related mean-risk models. SIAM J Optim 13(1):60–78
Perakis G, Roels G (2008) Regret in the newsvendor model with partial information. Oper Res 56(1):188–203
Petruzzi NC, Dada M (1999) Pricing and the newsvendor problem: a review with extensions. Oper Res 47(2):183–194
Pflug GC (2000) Some remarks on the value-at-risk and the conditional value-at-risk. In: Probabilistic constrained optimization: methodology and applications, pp. 272–281. Kluwer Academic Publishers, Dordrecht
Qiu R, Shang J, Huang X (2014) Robust inventory decision under distribution uncertainty: a CVaR-based optimization approach. Int J Prod Econ 153:13–23
Rockafellar RT, Uryasev S (2000) Optimization of conditional value-at-risk. J Risk 2(3):21–41
Rockafellar RT, Uryasev S (2002) Conditional value-at-risk for general loss distributions. J Bank Finance 26(7):1443–1471
Scarf HA (1958) A min–max solution of an inventory problem. In: Studies in the mathematical theory of inventory and production, pp. 201–209. Stanford University Press, California
Smith NR, Martinez-Flores JL, Cardenas-Barron LE (2007) Analysis of the benefits of joint price and order quantity optimisation using a deterministic profit maximisation model. Prod Plan Control 18(4):310–318
Song Y, Ray S, Boyaci T (2009) Optimal dynamic joint inventory-pricing control for multiplicative demand with fixed order costs and lost sales. Oper Res 57(1):245–250
Steiner WJ, Belitz C, Lang S (2006) Semiparametric stepwise regression to estimate sales promotion effects. In: Spiliopoulou M, Kruse R, Borgelt C, Nürnberger A, Gaul W (eds) From data and information analysis to knowledge engineering. Springer, Berlin, pp 590–597
Sundar DK, Ravikumar K (2018) The distribution free newsboy problem with partial information. Int J Oper Res 33(4):481–496
Wang CX, Webster S, Zhang S (2014) Robust price-setting newsvendor model with interval market size and consumer willingness-to-pay. Int J Prod Econ 154:100–112
Whitin TM (1955) Inventory control and price theory. Manag Sci 2(1):61–80
Xu M, Jianbin L (2010) Optimal decisions when balancing expected profit and conditional value-at-risk in newsvendor models. J Syst Sci Complex 23(6):1054–1070
Xue W, Ma L, Liu Y, Lin M (2022) Value of inventory pooling with limited demand information and risk aversion. Decis Sci 53(1):51–83
Yano CA, Gilbert SM (2005) Coordinated pricing and production/procurement decisions: a review. In: Chakravarty A, Eliashberg J (eds) Managing business interfaces: marketing, engineering and manufacturing perspectives. Kluwer Academic Publishers, Boston, pp 65–103
Yao L, Chen Y, Yan H (2006) The newsvendor problem with pricing: extensions. Int J Manag Sci Eng Manag 1(1):3–16
Zabel E (1972) Multiperiod monopoly under uncertainty. J Econ Theory 5(3):524–536
Zhou YJ, Chen XH, Wang ZR (2008) Optimal ordering quantities for multi-products with stochastic demand: return-CVaR model. Int J Prod Econ 112(2):782–795
Acknowledgements
The authors thank Referees for helpful comments on the paper. Sirong Luo’s research is supported by National Natural Science Foundation of China [NSFC-72271148].
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The authors have no competing interests to declare that are relevant to the content of this article.
Ethical approval
Ethics approval, Consent, Data, Materials and/or Code availability are not applicable to this article as no new data were created or analyzed in this study. All data in this paper are generated by numerical simulation.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendices
Appendix A: Derivations
Proof
1. The detailed derivation of the two-point cumulative distribution in Eqs. (15)–(18). We will show that there exists a two-point distribution \(D^{*}\thicksim \Phi ^{*}\in \mathscr {F}\) such that \(E\vert q-D\vert \le E\vert q-D^{*}\vert\) for all \(D\thicksim \Phi \in \mathscr {F}\).
First, the following proof will use the known fact that for any quadratic polynomial function H(x), we have \(E[H(D_1)]=E[H(D_2)]\) for all \(D_1,D_2\in \mathscr {F}\) (Scarf 1958). That is, if two random variables have the same expectation and the same variance, then their quadratic polynomial functions have the same expectation.
Second, we construct a quadratic polynomial function \(H_{0}(x)\), which satisfies \(H_{0}(x)\ge \vert q-x\vert\) for all \(x\ge 0\) and \(H_{0}(x)\) is tangent to \(\vert q-x\vert\) at two points \(x_1,x_2\), i.e., \(H_{0}(x_1)=\vert q-x_1\vert\) and \(H_{0}(x_2)=\vert q-x_2\vert\). It is easy to know that \(x_1,x_2\) are symmetric with respect to \(x=q\), i.e., there exists e such that \(x_1=q-e\) and \(x_2=q+e\).
Finally, define a two-point distribution \(D^{*}\), which assigns weight \(\beta\) to \(q-e\) and weight \(1-\beta\) to \(q+e\). It is easy to verify that \(e=(v^2+(q-z)^2)^{1/2}\) and \(\beta =\frac{(v^2+(q-z)^2)^{1/2}+(q-z)}{2(v^2+(q-z)^2)^{1/2}}\) to ensure \(E[D]=z,Var[D]=v^2\), i.e, \(D^{*}\thicksim \Phi ^{*}\in \mathscr {F}\).
Based on the above facts, we have \(E\vert q-D\vert \le E[H_{0}(D)]=E[H_{0}(D^{*})]=E\vert q-D^{*}\vert\). The first inequality is based on \(H_{0}(x)\ge \vert q-x\vert\) for all \(x\ge 0\), the second equality is based on \(D,D^{*}\in \mathscr {F}\), and the third equality is based on that \(D^{*}\) is a two-point distribution at \(x_1,x_2\) and \(H_{0}(x)\) is tangent to \(\vert q-x\vert\) at \(x_1,x_2\). \(\square\)
Proof
2. The detailed derivation of \(q^{*}(p)\)in Eq. (19). The first derivative of \(\pi (p,q)\) in Eq. (14):
The second derivative of \(\pi (p,q)\):
This means \(\pi (p,q)\) is concave in q. Thus, \(\frac{\partial \pi (p,q)}{\partial q}=0\) gives the optimal \(q^{*}(p)\) in Eq. (19). \(\square\)
Proof
3. The detailed derivation of \(\pi (p,q^{*}(p))\) in Eq. (20). Define \(U(p)=\frac{1}{1-(1-\alpha )(1-\frac{c}{p})}\), then \(q^{*}(p)=z+\frac{v(U-2)}{2\sqrt{U-1}}\) and
\(\square\)
Proof
4. The detailed derivation of Remark 1. \(\pi (p,q^{*}(p))\ge 0\) means \(p-c\ge \frac{\sigma ^2}{\mu ^2}\frac{1}{1-\alpha }(\alpha p+(1-\alpha )c)\), and then \((1-\frac{\sigma ^2}{\mu ^2}\frac{\alpha }{1-\alpha })p\ge (1+\frac{\sigma ^2}{\mu ^2})c\). So we have that \(1-\frac{\sigma ^2}{\mu ^2}\frac{\alpha }{1-\alpha }>0\) and \(p\ge \frac{(1+\frac{\sigma ^2}{\mu ^2})c}{1-\frac{\sigma ^2}{\mu ^2}\frac{\alpha }{1-\alpha }}\), they can be simplified as \(\alpha <\frac{\mu ^2}{\mu ^2+\sigma ^2}\) and \(p\ge (1+\frac{\sigma ^2}{\mu ^2-\alpha (\mu ^2+\sigma ^2)})c\). \(\square\)
Appendix B: Proofs
Proof
1. Proof of Theorem 1. Define \(X(p)=\sqrt{\frac{\alpha p+(1-\alpha )c}{(1-\alpha )(p-c)}}\frac{\sigma }{\mu }-1\), we have that X(p) is decreasing in p and \(X(p)\in (\sqrt{\frac{\alpha }{1-\alpha }}\frac{\sigma }{\mu }-1,0)\) for \((1+\frac{\sigma ^2}{\mu ^2-\alpha (\mu ^2+\sigma ^2)})c< p<\infty\), where \(\sqrt{\frac{\alpha }{1-\alpha }}\frac{\sigma }{\mu }-1\in [-1,0)\) for \(0\le \alpha <\frac{\mu ^2}{\mu ^2+\sigma ^2}\). The first derivative of X(p) is \(\frac{\partial X(p)}{\partial p}=-\frac{1}{2}\frac{1}{p-c}\frac{c}{\alpha p+(1-\alpha )c}(X(p)+1)\). Using the new function X(p), we transfer \(\pi (p,q^{*}(p))\) to
The first derivative of \(\pi (p,q^{*}(p))\):
where \(W(p)=\eta (p)(1-\frac{c}{p})X(p)-(X(p)-\frac{1}{2}\frac{c}{\alpha p+(1-\alpha )c}(X(p)+1))\).
The second derivative:
In order to prove the unimodality of \(\pi (p,q^{*}(p))\), we consider the sign of \(\frac{\partial ^2\pi (p,q^{*}(p))}{\partial p^2}\mid _{\frac{\partial \pi (p,q^{*}(p))}{\partial p}=0}\), that is, the sign of \(\frac{\partial W(p)}{\partial p}\mid _{W(p)=0}\). The condition \(W(p)=0\) means the fact \(\eta (p)=\frac{1}{2}\frac{p}{p-c}[2-\frac{c}{\alpha p+(1-\alpha )c}(1+\frac{1}{X(p)})]\). Then
We have that \(\frac{\partial W(p)}{\partial p}\mid _{W(p)=0}<0\) for \(\frac{\partial \eta (p)}{\partial p}>0\), \(-1<X(p)<0\), \(\eta (p)>0\), \(\frac{\partial X(p)}{\partial p}<0\). Therefore, \(\pi (p,q^{*}(p))\) is unimodal in p, and \(p^{*}\) is obtained by \(\frac{\partial \pi (p,q^{*}(p))}{\partial p}=0\), that is \(\eta (p)=\frac{1}{2}\frac{p}{p-c}[2-\frac{c}{\alpha p+(1-\alpha )c}(1+\frac{1}{X(p)})]\triangleq g_{\alpha }(p)\), where \(g_{\alpha }(p)\) is defined in Eq. (23).
Define \(\gamma (p)=2-\frac{c}{\alpha p+(1-\alpha )c}(1+\frac{1}{X(p)})\), then \(g_{\alpha }(p)=\frac{1}{2}\frac{p}{p-c}\gamma (p)\). Since \(-1<X(p)<0\), \(1+\frac{1}{X(p)}<0\) and \(X(p)+1>0\), we have that \(\gamma (p)>0\) and
\(\frac{1}{2}\frac{p}{p-c}\) and \(\gamma (p)\) are both positive and decreasing in p. Thus, \(g_{\alpha }(p)\) is decreasing in p. Moreover, based on the fact that \(\frac{\partial ^2 (f_{1}(p)f_{2}(p))}{\partial p^2}=\frac{\partial ^2 f_{1}(p)}{\partial p^2}f_{2}(p)+f_{1}(p)\frac{\partial ^2 f_{2}(p)}{\partial p^2}+2\frac{\partial f_{1}(p)}{\partial p}\frac{\partial f_{2}(p)}{\partial p}\), the second derivative of \(g_{\alpha }(p)\) is \(\frac{\partial ^2 g_{\alpha }(p)}{\partial p^2}=\frac{c}{(p-c)^3}\gamma (p)+\frac{p}{2(p-c)}\frac{\partial ^2 \gamma (p)}{\partial p^2}-\frac{c}{(p-c)^2}\frac{\partial \gamma (p)}{\partial p}\). We have that \(\frac{\partial ^2 g_{\alpha }(p)}{\partial p^2}\) is positive, ie. \(g_{\alpha }(p)\) is convex in p, for \(\gamma (p)>0\), \(\frac{\partial \gamma (p)}{\partial p}<0\), and \(\frac{\partial ^2 \gamma (p)}{\partial p^2}=-2\frac{c\alpha }{(\alpha p+(1-\alpha )c)^3}[\alpha (1+\frac{1}{X(p)})-\frac{1}{2}\frac{c}{p-c}\frac{X(p)+1}{X(p)^2}]+\frac{c}{(\alpha p+(1-\alpha )c)^2}[-\frac{\alpha }{X(p)^2}\frac{\partial X(p)}{\partial p}+\frac{1}{2}\frac{c}{(p-c)^2}\frac{X(p)+1}{X(p)^2}+\frac{1}{2}\frac{c}{p-c}\frac{X(p)+2}{X(p)^3}\frac{\partial X(p)}{\partial p}]>0\). \(\square\)
Proof
2. Proof of Lemma 2. We first simplify \(g_{\alpha }(p)\) using \(h_{\alpha }(p)\), and then analyze the sign of \(\frac{\partial h_{\alpha }(p)}{\partial \alpha }\) for the different p.
where \(h_{\alpha }(p)=\frac{\sigma }{\mu }(\alpha +\frac{c}{p-c})-\sqrt{(1-\alpha )(\alpha +\frac{c}{p-c})}\). For a fixed p, the monotonicity of \(g_{\alpha }(p)\) in \(\alpha\) is related to the monotonicity of \(h_{\alpha }(p)\) in \(\alpha\).
The first derivative of \(h_{\alpha }(p)\) with respect to \(\alpha\):
The second derivative of \(h_{\alpha }(p)\) with respect to \(\alpha\):
This means \(h_{\alpha }(p)\) is convex in \(\alpha\). Next we will discuss the sign of \(\frac{\partial h_{\alpha }(p)}{\partial \alpha }\).
Define \(l(\alpha ,p)\) as \(l(\alpha ,p)=\frac{\partial h_{\alpha }(p)}{\partial \alpha }=\frac{\sigma }{\mu }-\frac{1}{2}(\sqrt{\frac{1-\alpha }{\alpha +\frac{c}{p-c}}}-\sqrt{\frac{\alpha +\frac{c}{p-c}}{1-\alpha }})\) is a function of \(\alpha\) and p, it is increasing in \(\alpha\) and decreasing in p, next we will derive the ranges of \(\alpha\) and p which satisfy \(l(\alpha ,p)>0\) and \(l(\alpha ,p)<0\) respectively. For a given \(\alpha\), \(l(\alpha ,p)\) is decreasing to the lower bound \(l(\alpha ,\infty )=\frac{\sigma }{\mu }-\frac{(1-\alpha )-\alpha }{2\sqrt{(1-\alpha )\alpha }}\) when p increases to \(\infty\). We first analyze the sign of the lower bound \(l(\alpha ,\infty )\) and then the sign of \(l(\alpha ,p)\). The lower bound is increasing in \(\alpha\), and it is negative when \(0\le \alpha <\frac{\delta }{1+\delta }\) and positive when \(\frac{\delta }{1+\delta }<\alpha <\alpha _{max}\), where \(\frac{\delta }{1+\delta }\) is the solution of \(l(\alpha ,\infty )=0\). Thus, the sign of \(l(\alpha ,p)\) is divided into the following two cases.
For \(\forall \alpha \in [0,\frac{\delta }{1+\delta })\), the lower bound \(l(\alpha ,\infty )<0\), and because \(l(\alpha ,p)\) decreases in p, there exists a unique solution \(p_0\), which is defined as \(p_0=(1+\frac{1}{(1-\alpha )\delta -\alpha })c\), satisfying \(l(\alpha ,p_0)=0\) (We know \(p_0\) is indeed in the range \([p_{min},\infty )\) for the fact that \(l(\alpha ,p_{min})=\frac{1}{2}(\frac{\sigma }{\mu }+\frac{\mu }{\sigma })>0\), where \(p_{min}=(1+\frac{\sigma ^2}{\mu ^2-\alpha (\mu ^2+\sigma ^2)})c\)). For \(p> p_{0}\), we have that \(l(\alpha ,p)<0\) and \(g_{\alpha }(p)\) decreases in \(\alpha\); for \(p\le p_{0}\), we have that \(l(\alpha ,p)\ge 0\) and \(g_{\alpha }(p)\) increases in \(\alpha\).
For \(\forall \alpha \in [\frac{\delta }{1+\delta },\alpha _{max})\), the lower bound \(l(\alpha ,\infty )\ge 0\), and because \(l(\alpha ,p)\) decreases in p, the first order condition \(l(\alpha ,p)=0\) has no solution. In this case, \(l(\alpha ,p)>l(\alpha ,\infty )\ge 0\) and \(g_{\alpha }(p)\) is increasing in \(\alpha\) for all \(p\in [(1+\frac{\sigma ^2}{\mu ^2-\alpha (\mu ^2+\sigma ^2)})c,\infty )\). \(\square\)
Proof
3. Proof of Theorem 3. The outline of the proof is the following:
Part 1: We analyze the sensitivity of \(p^{*}(\alpha )\). Based on the results in Lemma 2, the analysis is divided into two ranges: \([0,\frac{\delta }{1+\delta })\) and \([\frac{\delta }{1+\delta },\alpha _{max})\). For \(\alpha \in [\frac{\delta }{1+\delta },\alpha _{max})\), \(g_{\alpha }(p)\) is increasing in \(\alpha\) for all p, thus, \(p^{*}(\alpha )\) is increasing in \(\alpha\). For \(\alpha \in [0,\frac{\delta }{1+\delta })\), \(g_{\alpha }(p)\) is increasing in \(\alpha\) for \(p\le p_{0}(\alpha )\) and decreasing in \(\alpha\) for \(p>p_{0}(\alpha )\), we need to compare \(p^{*}(\alpha )\) and \(p_0(\alpha )\), which is equivalent to compare \(g_{\alpha }(p_0(\alpha ))\) and \(\eta (p_0(\alpha ))\). For the \(\alpha\) satisfying that \(g_{\alpha }(p_0(\alpha ))<\eta (p_0(\alpha ))\), it satisfies that \(p^{*}(\alpha )<p_0(\alpha )\) and \(p^{*}(\alpha )\) is increasing, and vice versa.
Part 2: We analyze the sensitivity of \(q^{*}(\alpha )\) according to Eq. (24). To analyze the monotonicity of \((1-\alpha )(1-\frac{c}{p^{*}})\), we introduce a variable \(G^{*}(\alpha )\) which is a function of \(\alpha\) and \(p^{*}(\alpha )\), and we rewrite \(q^{*}(\alpha )\) as the function of \(y(p^{*})\) and \(G^{*}(\alpha )\), then analyze the monotonicity of \(G^{*}\) as \(\alpha\) increases.
Part 1: The sensitivity of \(p^{*}(\alpha )\).
First, we analyze the sensitivity of \(p^{*}(\alpha )\), which is the unique intersection of decreasing function \(g_{\alpha }(p)\) and increasing function \(\eta (p)\).
The monotonicity of \(g_{\alpha }(p)\) in \(\alpha \in [0,\alpha _{max})\) is divided into two ranges: \([0,\frac{\delta }{1+\delta })\) and \([\frac{\delta }{1+\delta },\alpha _{max})\).
For \(\alpha \in [\frac{\delta }{1+\delta },\alpha _{max})\), \(g_{\alpha }(p)\) is increasing in \(\alpha\) for all p and then \(p^{*}(\alpha )\) is increasing in \(\alpha\), which is according to Theorem 1 and Lemma 2.
For \(\alpha \in [0,\frac{\delta }{1+\delta })\), \(g_{\alpha }(p)\) is increasing in \(\alpha\) for \(p\le p_{0}(\alpha )\) and decreasing in \(\alpha\) for \(p>p_{0}(\alpha )\), according to Lemma 2. In order to analyze the sensitivity of \(p^{*}(\alpha )\), we need to compare \(p^{*}(\alpha )\) and \(p_0(\alpha )\). If \(p^{*}(\alpha )\le p_0(\alpha )\), then \(g_{\alpha }(p)\) is increasing in \(\alpha\) at all points in range \(p\in (p_{min},p_0(\alpha ))\) including \(p=p^{*}(\alpha )\), which implies \(p^{*}(\alpha )\) is increasing because it is the unique intersection of decreasing function \(g_{\alpha }(p)\) and increasing function \(\eta (p)\). Similarly, if \(p^{*}(\alpha )>p_0(\alpha )\), then \(g_{\alpha }(p)\) is decreasing in \(\alpha\) at all points in range \(p\in (p_0(\alpha ),\infty )\) including \(p=p^{*}(\alpha )\), which implies \(p^{*}(\alpha )\) is decreasing in \(\alpha\).
To compare \(p^{*}(\alpha )\) and \(p_0(\alpha )\), we need to analyze the solutions of \(p^{*}(\alpha )=p_0(\alpha )\), which is equivalent to \(g_{\alpha }(p_0(\alpha ))=\eta (p_0(\alpha ))\), i.e., \(g_{\alpha }((1+\frac{1}{(1-\alpha )\delta -\alpha })c)=\eta ((1+\frac{1}{(1-\alpha )\delta -\alpha })c)\). Specifically, \(g_{\alpha }((1+\frac{1}{(1-\alpha )\delta -\alpha })c)=k_1 \alpha +k_2\), where \(k_1=-\frac{(1+\delta )^2}{2\delta }<0\) and \(k_2=\frac{3+\delta }{2}>0\) are constants. Therefore, \(g_{\alpha }((1+\frac{1}{(1-\alpha )\delta -\alpha })c)\) decreases from \(k_2\) to \(k_1\frac{\delta }{1+\delta }+k_2\) when \(\alpha\) increases from 0 to \(\frac{\delta }{1+\delta }\). Note that \(\eta ((1+\frac{1}{(1-\alpha )\delta -\alpha })c)\) is increasing in \(\alpha\) from \(\eta (c+\frac{c}{\delta })\) to \(\infty\) when \(\alpha\) increases from 0 to \(\frac{\delta }{1+\delta }\).
If \(\eta (c+\frac{c}{\delta })\ge k_2\), then \(g_{\alpha }(p_0(\alpha ))\le \eta (p_0(\alpha ))\) for all \(\alpha \in [0,\frac{\delta }{1+\delta })\). Note \(p=p^{*}(\alpha )\) is the value which satisfies \(g_{\alpha }(p)=\eta (p)\). Therefore, we have that \(p^{*}(\alpha )\le p_0(\alpha )\) and then \(p^{*}(\alpha )\) is increasing in \(\alpha\) for all \(\alpha \in [0,\frac{\delta }{1+\delta })\).
If \(\eta (c+\frac{c}{\delta })< k_2\), then there exists a unique \(\alpha ^{*}\in [0,\frac{\delta }{1+\delta })\) which is the solution of
For \(0\le \alpha <\alpha ^{*}\), we have that \(g_{\alpha }(p_0(\alpha ))>\eta (p_0(\alpha ))\), which implies \(p^{*}(\alpha )> p_0(\alpha )\) and then \(p^{*}(\alpha )\) is decreasing in \(\alpha\). Similarly, for \(\alpha ^{*}\le \alpha <\frac{\delta }{1+\delta }\), we have that \(p^{*}(\alpha )\) is increasing in \(\alpha\).
Part 2: The sensitivity of \(q^{*}(\alpha )\).
Define \(G=(1-\alpha )(1-\frac{c}{p})\), and then \(p=\frac{c}{1-\frac{G}{1-\alpha }}\), \(\frac{p}{p-c}=\frac{1-\alpha }{G}\), \(\alpha +(1-\alpha )\frac{c}{p}=1-G\) and \(\alpha p+(1-\alpha )c=\frac{c}{1-\frac{G}{1-\alpha }}(1-G)\). Since \(p^{*}\) is the solution of \(g_{\alpha }(p)=\eta (p)\), so \(G^{*}=(1-\alpha )(1-\frac{c}{p^{*}})\) is the solution of \(g_{\alpha }(\frac{c}{1-\frac{G}{1-\alpha }})=\eta (\frac{c}{1-\frac{G}{1-\alpha }})\). Define \(\hat{g}_{\alpha }(G)=g_{\alpha }(\frac{c}{1-\frac{G}{1-\alpha }})\) and \(\hat{\eta }_{\alpha }(G)=\eta (\frac{c}{1-\frac{G}{1-\alpha }})\). Specifically,
Note that \(\frac{\partial \hat{g}_{\alpha }(G)}{\partial G}=\frac{\partial g(p)}{\partial p}\frac{\partial p}{\partial G}<0\), \(\hat{g}_{\alpha }(G)\) is a decreasing function of G. And for a fixed G, it is a decreasing function of \(\alpha\), for \(1+\frac{1}{\sqrt{\frac{1-G}{G}}\frac{\sigma }{\mu }-1}=1+\frac{1}{X(p)}<0\), where X(p) is defined in the proof of Theorem 1. Since \(\hat{\eta }_{\alpha }(G)=\eta (\frac{c}{1-\frac{G}{1-\alpha }})\), thus, \(\hat{\eta }_{\alpha }(G)\) is an increasing function of G, and for a fixed G, it is increasing in \(\alpha\).
In conclusion, when \(\alpha\) increases, we have that \(\hat{g}_{\alpha }(G)\) decreases for all G and \(\hat{\eta }_{\alpha }(G)\) increases for all G, therefore \(G^{*}\), the solution of \(\hat{g}_{\alpha }(G)=\hat{\eta }_{\alpha }(G)\), is decreasing in \(\alpha\) (see Fig. 5).
If \(p^{*}(\alpha )\) increases in \(\alpha\), then
decreases in \(\alpha\), for y(p) decreases in p, \(\frac{\frac{1}{1-G^{*}}-2}{\sqrt{\frac{1}{1-G^{*}}-1}}\) increases in \(G^{*}\) and \(G^{*}\) decreases in \(\alpha\). \(\square\)
Proof
4. Proof of Example 1. According to Theorem 3, there exists a unique \(\alpha ^{*}\in [0,\frac{\delta }{1+\delta })\) which is the solution of
Equation (B13) can be written as a quadratic equation with respect to \(\alpha\). Its roots are \(\frac{\pm \sqrt{(1+b^{2}c^2)\delta ^2+2bc\delta }+(\delta +2-bc)\delta }{(1+\delta )^2}\), the existence and uniqueness of \(\alpha ^{*}\in [0,\frac{\delta }{1+\delta })\) which is proved in Theorem 3 guarantees the value in the square root is positive. Moreover, the larger root satisfies
such that it is infeasible. Therefore, \(\alpha ^{*}=\frac{-\sqrt{(1+b^{2}c^2)\delta ^2+2bc\delta }+(\delta +2-bc)\delta }{(1+\delta )^2}\). \(\square\)
Proof
5. Proof of Theorem 4. Define \(B(p)=\sigma \sqrt{\frac{\alpha p+(1-\alpha )c}{(1-\alpha )(p-c)}}\), B(p) is decreasing in p and \(\frac{\partial B(p)}{\partial p}=-\frac{1}{2}\frac{1}{p-c}\frac{c}{\alpha p+(1-\alpha )c}B(p)\). Using the new function B(p), we transfer \(\pi (p,q^{*}(p))\) to
The first derivative of \(\pi (p,q^{*}(p))\):
Based the fact that \(y(p)\ge B(p)\) and \(1-\frac{1}{2}\frac{c}{\alpha p+(1-\alpha )c}\ge 0\), we assume that \(0\le 1-(1-\frac{c}{p})\eta (p) \le 1-\frac{1}{2}\frac{c}{\alpha p+(1-\alpha )c}\), i.e.,
to ensure there exists a solution of \(\frac{\partial \pi (p,q^{*}(p))}{\partial p}=0\). It implies that \(p\in [p_1, p_2]\), where \(p_1\) is the unique solution of \(\eta (p)=\frac{1}{2}\frac{c}{\alpha p+(1-\alpha )c}\frac{p}{p-c}\) and \(p_2\) is the unique solution of \(\eta (p)=\frac{p}{p-c}\). For \(p<p_1\), \(\frac{\partial \pi (p,q^{*}(p))}{\partial p}>0\) and \(\pi (p,q^{*}(p))\) is strictly increasing in p; for \(p>p_2\), \(\frac{\partial \pi (p,q^{*}(p))}{\partial p}<0\) and \(\pi (p,q^{*}(p))\) is strictly decreasing in p. Therefore, we only need to prove the unimodality property of \(\pi (p,q^{*}(p))\) for \(p\in [p_1, p_2]\), and the unique maximizer will be the global optimum solution.
The second derivative:
In order to prove the unimodality of \(\pi (p,q^{*}(p))\), we consider the sign of \(\frac{\partial ^2\pi (p,q^{*}(p))}{\partial p^2}\mid _{\frac{\partial \pi (p,q^{*}(p))}{\partial p}=0}\). The condition \(\frac{\partial \pi (p,q^{*}(p))}{\partial p}=0\) means the fact \(B(p)=[1-(1-\frac{c}{p})\eta (p)]\frac{\alpha p+(1-\alpha )c}{\alpha p+(1-\alpha )c-\frac{c}{2}}y(p)\). Then
The unimodality property of \(\pi (p,q^{*}(p))\) is proved and \(p^{*}\) satisfies \(\frac{\partial \pi (p,q^{*}(p))}{\partial p}=0\), i.e., \([1-(1-\frac{c}{p})\eta (p)]y(p)-(1-\frac{1}{2}\frac{c}{\alpha p+(1-\alpha )c})B(p)=0\). And then the optimal order quantity is \(q^{*}=y(p^{*})+\frac{\sigma (\frac{1}{1-(1-\alpha )(1-c/p^{*})}-2)}{2\sqrt{\frac{1}{1-(1-\alpha )(1-c/p^{*})}-1}}\) following Eq. (19).
Next, we will analyze how the optimal \(p^*\) changes with \(\alpha\), where \(p^*\) and \(\alpha\) satisfy \(\frac{\partial \pi (p,q^*(p))}{\partial p}=0\). Take the first derivative of the first-order condition \(\frac{\partial \pi (p,q^*(p))}{\partial p}=0\) with respect to \(\alpha\), we obtain
Since when \(p=p^*\), \(\frac{\partial ^2\pi (p,q^{*}(p))}{\partial p^2}\le 0\), and \(\frac{p(p-c)}{2(\alpha p+(1-\alpha )c)^2}B(P)\ge 0\), then we obtain \(\frac{dp^*}{d\alpha }\le 0\). \(\square\)
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Wang, W., Yang, Y. & Luo, S. The max–min newsvendor pricing problem under conditional value-at-risk criterion. Flex Serv Manuf J 36, 71–102 (2024). https://doi.org/10.1007/s10696-022-09472-9
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10696-022-09472-9